论文部分内容阅读
板形检测信息的模式分解是板形控制过程中的技术难点，该文提出的一种新的神经网络模式识别方法却可以解决这个难题。该识别方法的优点是：在ＡＲＴ网络的特征表示场中采用了具有正反馈和非线性变换的结构，能够有效地抑制板形检测数据中的干扰影响，提高了模式识别系统的抗干扰能力；在类别场中抛弃了传统的竞争学习机制，新的学习机制可以迅速分解板形模式；按照轧机执行机构板形控制的能力设置标准板形模式，可以对任意复杂形式的板形缺陷进行控制。用这种识别方法对实测板形进行了模式分解，识别结果完全正确，充分说明ＡＲＴ神经网络识别方法是一种理想的板形模式识别方法。
Pattern decomposition of flatness detection information is a technical difficulty in the process of shape control. A new neural network pattern recognition method proposed in this paper can solve this problem. The advantage of this method is that the structure of positive feedback and non-linear transformation is adopted in the feature representation field of ART network, which can effectively restrain the influence of interference in the plate detection data and improve the anti-interference ability of the pattern recognition system. Discarding the traditional competitive learning mechanism in the field of classes, the new learning mechanism can quickly decompose the plate pattern; and the standard plate pattern can be controlled according to the ability of the actuator of the rolling mill control system to control any complicated form of plate defect. With this method of identification, the shape of the measured plate is decomposed and the recognition result is completely correct, which fully shows that the ART neural network identification method is an ideal plate pattern recognition method.